Comping with gcc and cygwin? nvcc file.cu -ccbin /usr/bin/

Accelerated Computing CUDA CUDA Programming and Performance
1

I’ve been using cuda on linux for some time, and I’ve just gotten a new computer that comes with Windows Vista. I thought I would see if I can get cuda running. I do not have a copy of visual studio, and prefer to code in cygwin using xemacs as an editor.

However, now that I’ve installed the 2.0 beta driver, toolkit and sdk. I cannot compile. Here is the output I’m getting when I enter the command:

$ nvcc mc.cu -ccbin "c:\cygwin\bin"

nvcc fatal   : nvcc cannot find a supported cl version. Only MSVC 7.1 and MSVC 8.0 are supported

I get the same error if I run this from the windows command prompt

c:\Program Files\NVIDIA Corporation\NVIDIA CUDA SDK\projects\MCGPU>nvcc mc.cu -ccbin c:\cygwin\bin\

nvcc fatal   : nvcc cannot find a supported cl version. Only MSVC 7.1 and MSVC 8.0 are supported

So do have to use MSVC or is there another version of gcc I can use? Just need some clarity on the issue. Surely there is someone else who uses cygwin and gcc in order to compile their cuda code in Windows vista?

Regards,

Craig

2

You need the visual studio compiler.
This is mentioned in the release notes btw.

3

Thanks, but I don’t want to use the visual studio compiler. I did read the release notes. My question was can I use gcc instead, using the -ccbin option?

Craig

4

what part of ‘need’ did you not understand? ;)

I also do not want to use visual studio, but if you want to do CUDA on windows you need to.
You can download a free version that works with CUDA.

5

Well, i just read this post and i had the same question: Why is it not possible? Why is an option -ccbin included when it does not function?

BTW. I hate that visual studio environment ;-)

6

-ccbin can also be used to indicate that the Microsoft compiler is installed in a non-standard location.